CN104173030A - Pulse wave starting point real-time detection method resisting waveform change interference and application thereof - Google Patents

Abstract

The invention discloses a pulse wave starting point real-time detection method resisting waveform change interference and application of the method. The method is applied to the field of medical detection. The method includes the steps that human body pulse wave signals are acquired through a pulse wave sensing system and are filtered; first-order derivatives of the pulse wave signals are acquired, and first-order derivative signals are filtered; average slope values in front of and behind each point on first-order derivative waves are calculated according to a sliding window method, and trend change signals are acquired through subtraction of the average slope values; pulse wave starting points are defined as maximum values of the trend change signals in cardiac cycles. The trend change signals correspond to the change rates of radial vibration force applied to an artery tube wall when pulse waves reach measuring points physiologically, the maximum values of the trend change signals correspond to vibration force fastest change moments, namely, the pulse wave starting points. According to the method that the waveform trend change is calculated according to waveform information in sliding windows on the pulse waves, and pulse wave waveform deformation interference caused by motion, drug administration and other special states and noise and waveform baseline drift caused by mobile detection and other occasions can be effectively overcome.

Description

Pulse wave starting point real-time detection method and application thereof that a kind of anti-wave form varies is disturbed

Technical field

The present invention relates to pulse wave starting point real-time detection method and application that a kind of anti-wave form varies is disturbed, be applied in medical science detection field.

Background technology

Each cardiac cycle, the fast rapid fire blood of heart enters aorta, causes the flexible mechanical wave that produces of aorta tube wall, this mechanical wave along aorta tube wall towards branch and periphery blood vessel propagate, form pulse wave.Pulse wave originates in heart, and approach is tremulous pulse and histoorgan everywhere, buys property and gives birth to complicated echo at arterial branch, caliber variation and pipe wall material characteristic changing equipotential.Therefore, the pulse wave signal obtaining in parts of body measurement is the stack of heart incidence wave and flutter echo, is containing abundant cardiovascular system hemodynamics information.Measuring clinically pulse wave has had the history of upper a century, and doctors, by the analysis to pulse wave signal, extract various important physiological parameters, thereby know various physiology and the pathologic condition of cardiovascular system.The pulse wave signal obtaining at human body diverse location, is subject to the impact of echo stack, all has distortion in various degree on waveform, and centrifugal dirty distance pulse wave far away, is out of shape more serious.Initial point position is acknowledged as feature the most stable on pulse waveform, in pulse wave propagate process, not substantially being subject to the impact of echo stack, is the desired characteristics of calculating the important cardio-vascular parameters such as heart rate, change rate of heartbeat, pulse wave conduction speed, continuous blood pressure.Therefore,, in pulse wave analysis, the detection of starting point is most important.

Pulse wave starting point detects and is divided into off-line and the large class of real-time method two.Off-line checking method, as wavelet transformation, analysis of neural network etc., by near the feature of wave form varies complicated mathematic(al) manipulation or learning model extraction pulse wave starting point, can reach very high accuracy in detection conventionally.But these method computing complexity, amount of calculation is larger, is difficult to realize fast.Real-time detection method generally detects by the definition that the direct analysis of time-domain signal is provided to starting point, such as local minimum method (diastolic point), slope intersection method (tangent intersection) and second dervative maximum value process (maximum second derivative) etc.Amplitude minimum point on the legal adopted pulse wave of local minimum in from ECG signal R ripple position to this regional area of pulse wave first derivative maximum value position is pulse wave starting point.Specific practice is, by these regional area ten deciles, in each subregion, do smoothed curve matching and find out minimum point, 10 minimum point that again ten subregions obtained are respectively done smoothed curve matching again, the minimum point now obtaining is exactly the pulse wave starting point (list of references 1 of PPG, Xu P, Bergsneider M, Hu X (2009) Pulse onset detection using neighbor pulse-based signal enhancement.Med Eng Phys 31:337 – 345).Slope intersects method to be thought, the intersection point of the matching tangent line of the matching tangent line of pulse wave signal relaxing period descending branch and next cycle systole ascending branch is pulse wave starting point.Wherein, relaxing period descending branch is defined as the corresponding point of ECG signal R ripple on pulse wave and 2/5 cardiac cycle forward thereof, and systole ascending branch is pulse wave signal first derivative maximum of points and close region (list of references 2 thereof, Kazanavicius E, Gircys Rolandas, Vrubliauskas A (2005) Mathematical methods for determining the foot point of the arterial pulse wave and evaluation of proposed methods.Inform Technol Control 34:29 – 36).The acceleration maximum of points of second dervative maximum value process hypothesis pulse wave signal is starting point, utilize five dot center's difference formulas to obtain (list of references 2, Kazanavicius E, Gircys Rolandas, Vrubliauskas A (2005) Mathematical methods for determining the foot point of the arterial pulse wave and evaluation of proposed methods.Inform Technol Control 34:29 – 36).It is simple that the method for time domain possesses computing, and the feature that occupying system resources is little has the potentiality of extensive use in the pulse wave determination and analysis system of various portable, Wearables.But for actual measurement environment complicated and changeable, also there is certain defect in existing time domain approach.For example, show aspect the pulse waveform distortion that local minimum method and slope intersection method cause due to factors such as motion, medicines in opposing not good enoughly, second dervative maximum value process noise immunity is poor etc.

Based on this, we have proposed a kind of new time domain pulse wave starting point detection method, starting point is defined as to the maximum (maximum slope trend altering point, MSTA) of pulse wave trend variable signal, and has proposed corresponding automation algorithm.This method has embodied good real-time and accuracy in our experiment test.Compared with existing time domain detection method, the ability of this method opposing waveform distortion and baseline drift is especially outstanding.This method detects relevant all occasions by being expected to be applied to real-time pulse wave starting point, be particularly useful for the occasion that larger wave form varies and baseline drift appear in various pulse waves, the occasions such as the physiological and pathological states such as for example motion, administration and Aero-Space, portable detection.

Summary of the invention

The object of this invention is to provide a kind of peaked method of pulse wave trend variable signal that obtains in real time.

It is that this obtains in real time the peaked method of pulse wave trend variable signal and mainly comprises the steps: that the present invention realizes technical scheme that above-mentioned purpose takes

(1) gather pulse wave signal by pulse wave sensor-based system;

(2) pulse wave signal collecting is carried out to low-pass filtering;

(3) first derivative of calculating pulse wave signal;

(4) pulse wave first derivative signal is carried out to low-pass filtering;

(5) calculate the G-bar trend in the forward and backward side's sliding window of any point n on pulse wave first derivative signal

(6) calculate average tendency poor at any point n rear and front sliding window on pulse wave first derivative signal, acquisition trend variable signal

(7), at each cardiac cycle, extract trend variable signal maximum MSTA, be pulse wave starting point.

The peaked starting point that detects pulse wave that is applied as of pulse wave trend variable signal of the present invention.

The invention has the beneficial effects as follows: it is simple, easy to operate that (1) obtains the peaked method of pulse wave trend variable signal, can realize the real-time detection of pulse wave starting point; (2) when the corresponding pulse wave of pulse wave initial point position that method of the present invention detects arrives, radial vibration power changes the fastest moment, physiological significance definition clear-cut; (3) method of the present invention adopts sliding window calculating trend to change, thereby can effectively resist the impact that waveform distortion and baseline drift etc. are disturbed, be particularly suitable for being applied to the occasion such as the physiological and pathological states such as motion, administration and Aero-Space, portable detection.

Brief description of the drawings

Fig. 1. the pulse wave starting point feature that the present invention proposes: the computational methods key diagram of pulse wave trend variable signal maximum (MSTA).Sliding window width is defined as each W the point in impact point front and back.X ' is (n) any point on pulse wave first derivative signal, when calculate X ' (n) locate trend signal time, ask X ' (n) with W, its rear point X ' (n+1), X ' (n+2), X ' is (n+W) } the slope vector that forms respectively, these slope vector representations be taking X ' (n) as starting point, sensing X ' (n) W solid arrow of rear each point.The meansigma methods of these slope vectors is designated as be (n) starting point in order to X ', the empty arrow that points to its rear represents.Similarly, ask X ' (n) with W, its front point X ' (n-1), X ' (n-2), X ' is (n-W) } slope vector, these slope vectors for taking X ' (n) front each point as starting point, point to W solid arrow (n) of X '.The meansigma methods of these slope vectors is designated as use from X ' and (n) point to empty arrow (n) of X ' and represent in rear. deduct for the trend changing value that X ' (n) locates, on pulse wave first derivative signal trend changing value a little composition trend variable signal, and the maximum of trend variable signal in each cardiac cycle is pulse wave starting point;

Fig. 2. the oscillogram that each step of the inventive method draws, be followed successively by from top to bottom signal and slope trend variable signal after pulse wave signal after original pulse wave signal, filtering, pulse wave first derivative signal, first derivative signal filtering, wherein, "+" is pulse wave starting point corresponding point in each waveform;

Fig. 3. be respectively the principle comparison diagram of local minimum method, slope crossing method, second dervative maximum value process and the inventive method;

Fig. 4. the waveform change that motion causes intersects the impact of method testing result on slope.Figure (A1) and (B1) the photoelectricity volume pulse waveform that gathers under quiescent condition and kinestate respectively for the same tested object, wherein, ". " is artificial initial point position of demarcating, the starting point that "+" detected for the crossing method of slope.Figure (A2) and (B2) be schematic diagram, has represented respectively the testing result under the typical waveform of the crossing method of slope under tranquillization and kinestate.Wherein, two number lines represent the matched curve of relaxing period descending branch and systole ascending branch, and the position that two matched curve intersection points correspond on pulse waveform is starting point, represent with "+".Two vertical dotted line representatives in starting point left side start forwards 2/5 cardiac cycle from electrocardiosignal R ripple, and the vertical dotted line in two, right side represents pulse wave first derivative maximum and adjacent domain thereof.

Fig. 5. waveform change and the impact of baseline drift on local minimum method testing result that motion causes.Figure (A1) and (B1) the photoelectricity volume pulse waveform that gathers under quiescent condition and kinestate for the same tested object, wherein, ". " be the artificial initial point position of demarcation, "+" is the starting point of local minimum method detection.Figure (A2) and (B2) be schematic diagram, has represented respectively the testing process under the typical waveform of local minimum method under tranquillization and kinestate.Wherein, vertically dotted line representative is by electrocardiosignal R ripple position and the common starting point of delimiting of pulse wave first derivative maximum value position regional area around, round dot represents that regional area is divided into ten sections of matched curve minima of every section afterwards, and "+" is the minima of the matched curve of all round dots formation.

Fig. 6. the starting point testing result contrast of method of the present invention and other Time-Domain Detection Method.Wherein, (a) the one section of volume pulsation wave datum gathering under kinestate, (b), (c), (d) and (e) be followed successively by and adopt local minimum method, slope to intersect the pulse wave starting point testing result of method, second dervative maximum value process and the inventive method.Wherein, in (b), (c), (d), (e), the standard initial point position that ". " is manual detection, the testing result that "+" is automatic algorithms.

Detailed description of the invention

Below in conjunction with figure and implement the present invention is described further.

The present invention can obtain pulse wave trend variable signal maximum MSTA as follows in real time:

(1) adopt pulse wave sensor-based system to gather pulse wave signal.To gather the photoelectricity volume pulse wave signal of finger tip as example, photoelectricity volume pulse wave sensor is sandwiched in to finger tip, Real-time Collection volume pulsation wave signal, first carry out signal amplification and denoising by pulse wave signal Circuit tuning, be then converted to digital pulse ripple signal and be uploaded to PC by A/D change-over circuit.

(2) pulse wave signal collecting is carried out to low-pass filtering.To gather the photoelectricity volume pulse wave signal of finger tip as example, pulse wave signal is carried out to filtering with two rank low pass Butterworth filters, cut-off frequency is 15Hz.

(3) calculate the first derivative of original pulse wave signal.Calculate the first derivative of original pulse wave signal with forward difference formula, formula used is

x′(n)＝x _n-x _n-1,n＝1,2,3,...,N

(4) pulse wave first derivative signal is carried out to low-pass filtering.To gather the photoelectricity volume pulse wave signal of finger tip as example, adopt 50 rank low pass Finite Impulse Response filters to the signal filtering of photoelectricity volume pulsation wave first derivative, cut-off frequency is 20Hz.

(5) ask the slope trend variable signal in sliding window on pulse wave first derivative signal.

Any point n on pulse wave first derivative image, can be labeled as x'(n), as shown in Figure 1, at x'(n) in a front W neighborhood of a point, i point and x'(n) slope of line is:

$S_{-,i}\left(n\right)=\frac{X^{\′}\left(n\right)-X^{\′}(n-i)}{i},i=\mathrm{1,2},...W$

X'(n) with W, its front G-bar of putting the W bar straight line forming be:

${\stackrel{\‾}{S}}_{-}\left(n\right)=\frac{1}{W}\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\frac{X^{\′}\left(n\right)-X^{\′}(n-i)}{i}$

In the method, be used for characterizing x'(n) the slope trend in front.Similarly, at x'(n) in a rear W neighborhood of a point, i point and x'(n) slope of line is:

$S_{+,i}\left(n\right)=\frac{X^{\′}(n+i)-X^{\′}\left(n\right)}{i},i=\mathrm{1,2},...W$

In like manner can try to achieve an x'(n) with the G-bar of the W bar straight line of W, its rear point formation:

${\stackrel{\‾}{S}}_{+}\left(n\right)=\frac{1}{W}\underset{i=1}{\overset{W}{\mathrm{\Σ}}}\frac{X^{\′}(n+i)-X^{\′}\left(n\right)}{i}$

In the method, be used for characterizing x'(n) the slope trend at rear.The front and back trend variable signal of point n is defined as the difference of a n rear slope trend and front slope trend:

${\stackrel{\‾}{S}}_{\mathrm{diff}}\left(n\right)={\stackrel{\‾}{S}}_{+}\left(n\right)-{\stackrel{\‾}{S}}_{-}\left(n\right)$

W is defined as 1% of pulse wave signal sample rate.

(6) at each cardiac cycle, extract real-time slope trend variable signal maximum MSTA, MSTA position is defined as pulse wave starting point.We adopt dynamic bi-threshold method to obtain MSTA.The 2S that trend signal is started is most defined as the learning period, is two height thresholds (T1 and T2) initializes.T1 is defined as peaked half in learning period signal, and T2 is defined as the half of T1 value.Start anew subsequently to travel through trend variable signal, amplitude exceed T1 and thereafter the peak of ghost peak shape (crest is defined as the peak with certain rising and decline width) be defined as the maximum MSTA of current cardiac cycle trend variable signal.A new MSTA often detected, dynamically update the value of T1 and T2, formula is: T1_new=T1_old*0.75+T _mSTA* 0.5*0.25, T2_new=T1_new*0.5, wherein T _mSTArepresent the height of MSTA.New MSTA do not detected if the upper effective MSTA of distance exceedes in normal cardiac cycle scope (normal cardiac cycle scope is 0.3 second to 1.2 seconds), change current height criterion into threshold value T2 from threshold value T1, re-start searching.

The physiological significance of the pulse wave starting point detection method that the present invention proposes is explained as follows.From physiological angle, pulse wave starting point represents that each cardiac cycle propagates into the moment of pulse wave measurement point from the incidence wave of heart generation.In the time that incident mechanical wave is transferred to measuring point, the vibration force that incident mechanical wave causes vertically acts on measurement point ductus arteriosus wall, causes the motion of tube wall enlargement and contraction, forms pulse wave.And the moment (pulse wave starting point) that reaches of incidence wave is the fastest point of this vibration force variation.According to Newton's second law, F=ma, wherein, F is vibration force, and m is the quality of unit ductus arteriosus wall, is constant, and a is the acceleration that vibration force causes tube wall motion.Because m is constant, it is that corresponding acceleration a changes the fastest point that vibration force F changes the fastest point.Due to the second dervative of the corresponding pulse wave signal of acceleration a of tube wall motion, therefore, the maximum of the fastest corresponding pulse wave signal three order derivatives of point of acceleration change.From can find out, in the time of the neighborhood W=1 of a n, be the maximum of corresponding pulse wave three order derivatives, on physiological significance, vibration force changes the fastest point, and we are defined as the starting point of pulse wave.But, because the process to signal derivation is equivalent to high-pass filtering, can cause the decline of signal anti-noise ability, therefore, we set W is 1% of signal sampling rate, make feature that the present invention defines on the basis of original three order derivative maximum physiological significances, added the mean filter that W is ordered, in order to increase noiseproof feature.

Pulse wave trend variable signal maximum of the present invention can effectively be resisted the interference of waveform distortion and baseline drift, detects in real time, exactly the initial point position of pulse wave.In the real time signal processing of pulse wave, waveform distortion and baseline drift are two thorny interference factors.Waveform distortion mainly occurs in the situation that various factors causes cardiovascular system part or overall impedance characteristic to change.For example regulate the medicine of arteriotony can change the compliance of periphery blood vessel, and then change Peripheral resistance, thereby affect power and the spread speed of echo, cause pulse wave signal to produce distortion.Again for example under kinestate, motion causes the effect of health series of complex, as increasing local blood by expansion (decline of periphery peripheral vascular resistance), skeletal muscle peripheral vessels supplies with, improve rate of metabolism, thereby produce more energy for Skeletal Muscle Contraction, heart rate increases simultaneously, cardiac cycle shortens, the synthesis result of these factors causes the dicrotic wave amplitude of pulse wave to decline, and dicrotic wave position is to the skew of relaxing period afterbody, the distortion (as shown in Figure 4,5) that waveform generation is serious.Baseline drift is mainly caused by external disturbance, comprises action tested in signal measurement process, speaks, the shake of deep breathing, measuring point etc.As shown in Figure 3, existing local minimum method and slope intersect method in the time finding pulse wave starting point, first adopt electrocardiosignal R ripple and pulse wave first derivative maximum to locate near the neighborhood of starting point, then the intersection point of finding local minimum or waveform fitting straight line in this neighborhood, is defined as starting point.This neighborhood with electrocardiosignal R ripple location, there are two remarkable shortcomings, first be necessary synchronous acquisition electrocardiosignal, when making measurement more complicated, the implementation result of automatic algorithms is also subject to the impact of electrocardiosignal quality, these non-pulse wave factors of R ripple detection accuracy; Second is exactly that to resist waveform deformability poor.Fig. 4 has shown the continuous variation of one group of pulse wave from quiescent condition to kinestate waveform, we can see, along with the reduction of dicrotic wave height and the rising of relaxing period tail position, the fixing neighborhood of being located by electrocardiosignal R ripple cannot correctly have been followed the trail of the real trend of pulse wave starting point front end, thereby causes the cautious result that slope intersects method to produce larger error.Compared with fixing neighborhood, the trend variable signal calculating in the present invention is obtained by the sliding window dynamic calculation on first derivative signal completely, and represents that the trend variable signal maximum of initial point position is only calculated and obtained by the trend change information of (signal of front and back W width) in extremely narrow neighborhood around starting point.Even under larger wave form varies, this trend change information is also not easy to be affected.The major effect of baseline drift is the skew of pulse wave local minimum in the neighborhood that can cause being located by electrocardiosignal R ripple, causes the cautious result of local minimum method to produce wrong (as Fig. 5).Because method of the present invention is not defined as starting point feature by local minimum, but by calculating the variation of each some forward backward averaging slope vector, obtain because pulse wave arrives the local pulse waveform trend change information that measurement point causes, this information, from defining itself, just can not be subject to the impact of local minimum change in location.Simultaneously due to the sliding window dynamic calculation impact point trend change information in extremely narrow neighborhood around, waviness in is on a large scale affected to significant baseline drift, little to the trend variable effect that moving point is local.Therefore, the present invention also can suppress the impact of baseline drift well.In addition, because the process to signal derivation is equivalent to high-pass filtering, can cause the decline of signal anti-noise ability, make the anti-noise ability of second dervative maximum value process not good enough, and the present invention is in order to overcome this shortcoming, in computational process, add the mean filter that W is ordered, in order to increase noiseproof feature.

Using sliding window is the reason that the present invention is better than other three kinds of methods on the capacity of resisting disturbance that detects pulse wave starting point.Below using the photoelectricity volume pulsation wave data that gather under one section of kinestate as illustration:

Fig. 6 has shown the photoelectricity volume pulsation wave data (a figure) that gather under one section of kinestate, and adopts four kinds of time domain pulse wave starting point detection methods: local minimum method (b figure), slope intersect the testing result contrast of method (c figure), second dervative maximum value process (d figure) and the inventive method (e figure).As can be seen from Figure 4 and Figure 5, the pulse wave data that gather under kinestate have obvious baseline drift, synchronous signal waveform distortion serious (dicrotic wave height obviously reduces, and move to relaxing period afterbody position simultaneously).In the case, the testing result of the inventive method standard results artificial cautious with doctor kept highly consistent, is better than on the whole other three kinds of methods.

In sum, it is high, real-time that the method that the present invention proposes detects accuracy in the starting point of pulse wave signal, and the factors such as waveform distortion and baseline drift of can resisting are disturbed, and have broad application prospects in pulse wave real-time analysis field.

Claims (6)

1. the pulse wave starting point real-time detection method that anti-wave form varies is disturbed, is characterized in that, comprises the steps:

(1) gather pulse wave signal by pulse wave sensor-based system;

(2) pulse wave signal collecting is carried out to low-pass filtering;

(3) calculate the first derivative of original pulse wave signal;

(4) pulse wave first derivative signal is carried out to low-pass filtering;

(5) calculate the G-bar trend in the forward and backward side's sliding window of any point n on pulse wave first derivative signal

(6) the slope trend variable signal of any point n on calculating pulse wave first derivative ripple

(7), at each cardiac cycle, extract slope trend variable signal maximum MSTA, this maximum of points is the pulse wave starting point in this cardiac cycle;

2. pulse wave sensor-based system according to claim 1 gathers pulse wave signal, it is characterized in that: described pulse wave sensor-based system includes wound sensor-based system and noinvasive sensor-based system.

3. pulse wave sensor-based system according to claim 1 gathers pulse wave signal, it is characterized in that: described pulse wave signal comprises pressure pulse wave signal and volume pulsation wave signal.

4. pulse wave starting point according to claim 1, is characterized in that: described pulse wave trend variable signal maximum of points corresponds to pulse wave starting point.

5. the application of the pulse wave starting point real-time detection method that anti-wave form varies according to claim 1 is disturbed, it is characterized in that: described pulse wave starting point detects can effectively resist the impact that waveform distortion and baseline drift etc. are disturbed, and is applicable to the occasion such as the physiological and pathological states such as motion, administration and Aero-Space, portable detection.

6. the application of the pulse wave starting point real-time detection method that anti-wave form varies according to claim 1 is disturbed, it is characterized in that: it is simple that described pulse wave starting point detects the peaked method of acquisition pulse wave first derivative ripple slope trend variable signal, easy to operate, can realize the real-time detection of pulse wave starting point.